University of Twente Student Theses


Predicting influence spread in online social networks using combinations of node centralities

Praas, J.W. (2020) Predicting influence spread in online social networks using combinations of node centralities.

[img] PDF
Abstract:The extent and magnitude of information spread through an online social network is of great interest to marketing strategists and social scientists. The spread of information within a network has a strong correlation with the network statistics of the participating nodes. While the effects of individual node centralities on the influence spread has been researched by many, the predictive power of these centralities (or combinations thereof) have not. In this paper we show how well individual and combinations of those network statistics can predict the total estimated information spread (influence spread). We look at all non-isomorphic graphs of size N <= 9 nodes and small-to-medium graphs of N in {50, 100, 200, 400} nodes, randomly generated using the barabási-albert random graph generation model. Next, the Independent Cascade (IC) and Weighted Cascade (WC) spread models are used on each individual seed node in each graph to simulate the spread of information. Finally, the Machine Learning techniques Random Forest Regression and k-Nearest Neighbors regression are used to predict the total information spread. We find that the WC spread model results in higher R^2 scores than the IC model. The combination of the centralities Degree and PageRank and Betweenness are particularly predictive of the influence spread, both for IC and WC.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:50 technical science in general, 54 computer science, 70 social sciences in general
Programme:Computer Science BSc (56964)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page